All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2 1/4] drm: Mark up accesses of vblank->enabled outside of its spinlock
@ 2017-03-17 20:20 Chris Wilson
  2017-03-17 20:20 ` [PATCH v2 2/4] drm: vblank cannot be enabled if dev->irq_enabled is false Chris Wilson
                   ` (5 more replies)
  0 siblings, 6 replies; 11+ messages in thread
From: Chris Wilson @ 2017-03-17 20:20 UTC (permalink / raw)
  To: dri-devel; +Cc: Daniel Vetter, intel-gfx

Order the update to vblank->enabled after the timestamp is primed so
that a concurrent unlocked reader will only see the vblank->enabled with
the current timestamp.

v2: vblank->enable is guarded by dev->vbl_lock not
dev->vblank_time_lock, update the READ_ONCE accordingly.

Do not add a READ_ONCE(vblank->enabled) inside the interrupt handler to
avoid missing an interrupt whilst racing with enable_vblank()

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/drm_irq.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
index 53a526c7b24d..c47e07c89136 100644
--- a/drivers/gpu/drm/drm_irq.c
+++ b/drivers/gpu/drm/drm_irq.c
@@ -325,6 +325,8 @@ static void vblank_disable_and_save(struct drm_device *dev, unsigned int pipe)
 	struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
 	unsigned long irqflags;
 
+	assert_spin_locked(&dev->vbl_lock);
+
 	/* Prevent vblank irq processing while disabling vblank irqs,
 	 * so no updates of timestamps or count can happen after we've
 	 * disabled. Needed to prevent races in case of delayed irq's.
@@ -336,10 +338,8 @@ static void vblank_disable_and_save(struct drm_device *dev, unsigned int pipe)
 	 * calling the ->disable_vblank() operation in atomic context with the
 	 * hardware potentially runtime suspended.
 	 */
-	if (vblank->enabled) {
+	if (cmpxchg_relaxed(&vblank->enabled, true, false))
 		__disable_vblank(dev, pipe);
-		vblank->enabled = false;
-	}
 
 	/*
 	 * Always update the count and timestamp to maintain the
@@ -384,7 +384,7 @@ void drm_vblank_cleanup(struct drm_device *dev)
 	for (pipe = 0; pipe < dev->num_crtcs; pipe++) {
 		struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
 
-		WARN_ON(vblank->enabled &&
+		WARN_ON(READ_ONCE(vblank->enabled) &&
 			drm_core_check_feature(dev, DRIVER_MODESET));
 
 		del_timer_sync(&vblank->disable_timer);
@@ -1105,11 +1105,16 @@ static int drm_vblank_enable(struct drm_device *dev, unsigned int pipe)
 		 */
 		ret = __enable_vblank(dev, pipe);
 		DRM_DEBUG("enabling vblank on crtc %u, ret: %d\n", pipe, ret);
-		if (ret)
+		if (ret) {
 			atomic_dec(&vblank->refcount);
-		else {
-			vblank->enabled = true;
+		} else {
 			drm_update_vblank_count(dev, pipe, 0);
+			/* drm_update_vblank_count() includes a wmb so we just
+			 * need to ensure that the compiler emits the write
+			 * to mark the vblank as enabled after the call
+			 * to drm_update_vblank_count().
+			 */
+			WRITE_ONCE(vblank->enabled, true);
 		}
 	}
 
@@ -1517,7 +1522,7 @@ static int drm_queue_vblank_event(struct drm_device *dev, unsigned int pipe,
 	 * vblank disable, so no need for further locking.  The reference from
 	 * drm_vblank_get() protects against vblank disable from another source.
 	 */
-	if (!vblank->enabled) {
+	if (!READ_ONCE(vblank->enabled)) {
 		ret = -EINVAL;
 		goto err_unlock;
 	}
@@ -1644,7 +1649,7 @@ int drm_wait_vblank(struct drm_device *dev, void *data,
 		DRM_WAIT_ON(ret, vblank->queue, 3 * HZ,
 			    (((drm_vblank_count(dev, pipe) -
 			       vblwait->request.sequence) <= (1 << 23)) ||
-			     !vblank->enabled ||
+			     !READ_ONCE(vblank->enabled) ||
 			     !dev->irq_enabled));
 	}
 
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 11+ messages in thread

end of thread, other threads:[~2017-03-29 11:32 UTC | newest]

Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-03-17 20:20 [PATCH v2 1/4] drm: Mark up accesses of vblank->enabled outside of its spinlock Chris Wilson
2017-03-17 20:20 ` [PATCH v2 2/4] drm: vblank cannot be enabled if dev->irq_enabled is false Chris Wilson
2017-03-17 20:20 ` [PATCH v2 3/4] drm: Refactor vblank sequence number comparison Chris Wilson
2017-03-22  8:54   ` Michel Dänzer
2017-03-22 10:06   ` [PATCH v2] " Chris Wilson
2017-03-23  3:23     ` Michel Dänzer
2017-03-29 11:32       ` Ville Syrjälä
2017-03-17 20:20 ` [PATCH v2 4/4] drm: Peek at the current counter/timestamp for vblank queries Chris Wilson
2017-03-17 20:37 ` ✓ Fi.CI.BAT: success for series starting with [v2,1/4] drm: Mark up accesses of vblank->enabled outside of its spinlock Patchwork
2017-03-17 20:44 ` [PATCH v2 1/4] " Ville Syrjälä
2017-03-22 10:37 ` ✓ Fi.CI.BAT: success for series starting with [v2,1/4] drm: Mark up accesses of vblank->enabled outside of its spinlock (rev2) Patchwork

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.